A fuzzy compensation method, device, domain controller, and vehicle

Abstract

This application provides a blur compensation method, apparatus, domain controller, and vehicle, belonging to the field of automotive technology. The method includes: linearly fusing highlight pixel data and dark pixel data acquired by an image sensor of an onboard camera to obtain a base image; extracting a sharp component from the base image based on the vehicle's motion parameters; and then performing blur compensation on the highlight and dark pixel data based on the sharp component to determine a fused image based on the blurred highlight and dark pixel data. This method compensates for motion blur in the highlight and dark pixel data by using the sharp component to recover the edge structure and texture information contaminated by motion blur in the final fused image. This provides a reliable visual image for the perception and decision-making capabilities of autonomous driving systems, ensuring the effectiveness of autonomous driving systems and improving vehicle safety.

Description

Technical Field

[0001] This application relates to the field of automotive technology, and in particular to a fuzzy compensation method, device, domain controller, and vehicle. Background Technology

[0002] Vehicles often encounter HDR (High Dynamic Range) scenes where bright and dark areas coexist, such as driving against the light. In HDR scenes, due to the limited light recording range of the image sensor in the vehicle camera, it is difficult to capture the details of both bright and dark areas in the same frame.

[0003] To obtain an HDR image that simultaneously contains details from two different regions, the technology incorporates two types of pixel units with different light-sensing capabilities in the image sensor. This allows for the separate acquisition of pixel data for highlight and shadow areas within a single exposure, and the HDR image is obtained by fusing the two types of pixel data.

[0004] However, if pixel data becomes blurry, the blurry areas will be superimposed and amplified during fusion, resulting in visual defects such as color shift in HDR images. This affects the perception and decision-making capabilities of autonomous driving systems, thereby impacting vehicle safety. Summary of the Invention

[0005] This application provides a blur compensation method, apparatus, domain controller, and vehicle, capable of compensating for motion blur in highlight and shadow pixel data. This provides reliable visual images for the perception and decision-making capabilities of autonomous driving systems, ensuring the effectiveness of autonomous driving systems and improving vehicle safety. The technical solution is as follows: On the one hand, a fuzzy compensation method is provided, which includes: The highlight pixel data and dark pixel data collected by the image sensor of the vehicle's on-board camera are linearly fused to obtain the base image; Extract sharp components from the base image based on the vehicle's motion parameters; Based on the sharpness component, blur compensation is performed on the highlight pixel data and the shadow pixel data, and the fused image is determined based on the blurred highlight pixel data and shadow pixel data.

[0006] In some embodiments, extracting sharp components from a base image based on vehicle motion parameters includes: Determine the motion blur intensity of the base image based on the vehicle's motion parameters; Extracting sharp components from the base image based on motion blur intensity.

[0007] In some embodiments, determining the motion blur intensity of the base image based on the vehicle's motion parameters includes: Based on the vehicle's motion parameters, the blur length and blur direction of the base image are determined; The basic blur intensity of each pixel is determined based on the gradient magnitude of each pixel in the base image. Based on the gradient direction, blur length, and blur direction of each pixel, the basic blur intensity is corrected to obtain the motion blur intensity of each pixel.

[0008] In some embodiments, extracting the sharp component from the base image based on the motion blur intensity includes: The frequency domain transformation of the base image is performed based on the motion blur intensity to obtain the image spectrum of the base image; Filter the image spectrum to obtain the sharp component spectrum of the base image; The sharp component spectrum is inversely transformed in the frequency domain to obtain the sharp component of each pixel in the base image.

[0009] In some embodiments, frequency domain transformation of the base image is performed based on motion blur intensity to obtain the image spectrum of the base image, including: Based on the motion blur intensity of each pixel in the base image, the original gray values ​​of each pixel are weighted to obtain the blurred modulated gray values ​​of each pixel. The image spectrum of the base image is obtained by performing a Fourier transform on the blurred modulated gray values ​​of all pixels in the base image.

[0010] In some embodiments, after determining the blur length and blur direction of the base image based on the vehicle's motion parameters, the method further includes: Construct a directional filtering function based on the fuzzy length and fuzzy direction; Filtering the image spectrum yields the sharp component spectrum of the underlying image, including: Directional filtering of the image spectrum is performed using a directional filtering function to obtain the sharp component spectrum of the base image.

[0011] In some embodiments, blur compensation is performed on highlight pixel data and shadow pixel data based on the sharpness component to determine a fused image based on the blurred highlight pixel data and shadow pixel data, including: The sharp component, highlight pixel data, and shadow pixel data are fused to obtain a fused image.

[0012] In some embodiments, after performing an inverse frequency domain transformation on the sharp component spectrum to obtain the sharp component of each pixel in the base image, the method further includes: The motion blur intensity of each pixel is determined as the first fusion weight of the sharp component of each pixel; The sharp component, highlight pixel data, and shadow pixel data are fused to obtain a fused image, including: The second fusion weight is obtained by acquiring the first gray value of each pixel in the highlight pixel data, and the third fusion weight is obtained by acquiring the second gray value of each pixel in the dark pixel data. Based on the first fusion weight, the second fusion weight, and the third fusion weight, the sharp component, the first gray value, and the second gray value of each pixel are weighted and fused to obtain each fused pixel. A fused image is generated based on each fused pixel.

[0013] In some embodiments, obtaining a second fusion weight for the first grayscale value of each pixel in the highlight pixel data, and obtaining a third fusion weight for the second grayscale value of each pixel in the dark pixel data, includes: Obtain the basic fusion weight of the first gray value of each pixel in the highlight pixel data, and obtain the basic fusion weight of the second gray value of each pixel in the dark pixel data. Based on the motion blur intensity and sharpness component of each pixel, the basic fusion weights of the first grayscale value and the second grayscale value are modified respectively to obtain the second fusion weight of the first grayscale value and the third fusion weight of the second grayscale value.

[0014] In some embodiments, obtaining the basic fusion weight of the first grayscale value of each pixel in the highlight pixel data and obtaining the basic fusion weight of the second grayscale value of each pixel in the dark pixel data includes: Based on the saturation response value of the image sensor, the first gray value of each pixel in the highlight pixel data and the second gray value in the dark pixel data are normalized to obtain the normalized gray value of each pixel. Based on the normalized gray values ​​of each pixel, the basic fusion weights of the first gray value and the second gray value are determined.

[0015] On the other hand, a fuzzy compensation device is provided, the device comprising: The linear fusion unit is configured to linearly fuse the highlight pixel data and dark pixel data collected by the image sensor of the vehicle-mounted camera to obtain a base image. The component extraction unit is configured to extract sharp components from the base image based on the vehicle's motion parameters; The blur compensation unit is configured to perform blur compensation on highlight pixel data and shadow pixel data based on the sharp component, so as to determine the fused image based on the blur-compensated highlight pixel data and shadow pixel data.

[0016] In some embodiments, the component extraction unit includes: a blur intensity determination subunit, configured to determine the motion blur intensity of the base image based on the vehicle's motion parameters; The sharp component extraction subunit is configured to extract sharp components from the base image based on motion blur intensity.

[0017] In some embodiments, the blur intensity determination subunit is configured to determine the blur length and blur direction of the base image based on the vehicle's motion parameters; The basic blur intensity of each pixel is determined based on the gradient magnitude of each pixel in the base image. Based on the gradient direction, blur length, and blur direction of each pixel, the basic blur intensity is corrected to obtain the motion blur intensity of each pixel.

[0018] In some embodiments, the sharp component extraction subunit is configured to perform frequency domain transformation on the base image based on the motion blur intensity to obtain the image spectrum of the base image; Filter the image spectrum to obtain the sharp component spectrum of the base image; The sharp component spectrum is inversely transformed in the frequency domain to obtain the sharp component of each pixel in the base image.

[0019] In some embodiments, the sharp component extraction subunit is configured to weight the original gray value of each pixel based on the motion blur intensity of each pixel in the base image to obtain the blurred modulated gray value of each pixel. The image spectrum of the base image is obtained by performing a Fourier transform on the blurred modulated gray values ​​of all pixels in the base image.

[0020] In some embodiments, the fuzziness intensity determination subunit is configured to construct a directional filtering function based on the fuzziness length and the fuzziness direction; The sharp component extraction subunit is configured to perform directional filtering on the image spectrum based on a directional filtering function to obtain the sharp component spectrum of the base image.

[0021] In some embodiments, the blur compensation unit includes a fusion subunit configured to fuse the sharp component, highlight pixel data, and dark pixel data to obtain a fused image.

[0022] In some embodiments, the sharp component extraction subunit is configured to determine the motion blur intensity of each pixel as the first fusion weight of the sharp component of each pixel. The fusion subunit is configured to obtain a second fusion weight for each pixel in the highlight pixel data and a third fusion weight for each pixel in the dark pixel data. Based on the first fusion weight, the second fusion weight, and the third fusion weight, the sharp component, the first gray value, and the second gray value of each pixel are weighted and fused to obtain each fused pixel. A fused image is generated based on each fused pixel.

[0023] In some embodiments, the fusion subunit is configured to obtain the basic fusion weight of the first gray value of each pixel in the highlight pixel data and the basic fusion weight of the second gray value of each pixel in the dark pixel data. Based on the motion blur intensity and sharpness component of each pixel, the basic fusion weights of the first grayscale value and the second grayscale value are modified respectively to obtain the second fusion weight of the first grayscale value and the third fusion weight of the second grayscale value.

[0024] In some embodiments, the fusion subunit is configured to normalize the first gray value of each pixel in the highlight pixel data and the second gray value in the dark pixel data according to the saturation response value of the image sensor, so as to obtain the normalized gray value of each pixel. Based on the normalized gray values ​​of each pixel, the basic fusion weights of the first gray value and the second gray value are determined.

[0025] On the other hand, a domain controller is provided, which includes a main control module, the main control module including a processor and a memory, the memory being used to store at least one computer program, the at least one computer program being loaded and executed by the processor to implement the fuzzy compensation method in the embodiments of this application.

[0026] On the other hand, a vehicle is provided, which includes a domain controller for performing the fuzzy compensation method in the embodiments of this application.

[0027] On the other hand, a computer-readable storage medium is provided for storing at least one computer program, which is loaded and executed by a processor to implement the fuzzy compensation method in the embodiments of this application.

[0028] On the other hand, a computer program product is provided, including a computer program that is executed by a processor to implement the fuzzy compensation method in the embodiments of this application.

[0029] This application provides a blur compensation method that extracts a sharp component, i.e., the original structural information that is not contaminated by motion blur, from a base image using motion parameters, and compensates for motion blur in highlight pixel data and shadow pixel data using the sharp component.

[0030] The technical effect of this application embodiment is that it ensures that the final fused image can not only take into account the details of the highlight area and the dark area, but also restore the edge structure and texture information contaminated by motion blur, providing a reliable visual image for the perception and decision-making capabilities of the autonomous driving system, ensuring the effectiveness of the autonomous driving system and improving vehicle safety. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a schematic diagram of the implementation environment of a fuzzy compensation method provided in the embodiments of this application; Figure 2 This is a flowchart of a fuzzy compensation method provided according to an embodiment of this application; Figure 3 This is a flowchart of another fuzzy compensation method provided according to an embodiment of this application; Figure 4 This is a block diagram of a fuzzy compensation device according to an embodiment of this application; Figure 5 This is a schematic diagram of the structure of a domain controller according to an embodiment of this application. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0034] In this application, the terms "first", "second", etc. are used to distinguish identical or similar items with essentially the same function. It should be understood that there is no logical or temporal dependency between "first", "second", and "nth", nor is there any limitation on the quantity or execution order.

[0035] In this application, the term "at least one" means one or more, and "multiple" means two or more.

[0036] It should be noted that all information, data and signals involved in this application are authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0037] Figure 1 This is a schematic diagram illustrating the implementation environment of a fuzzy compensation method according to an embodiment of this application. See also... Figure 1 The implementation environment is in a vehicle 100, which is equipped with an onboard camera 101, an IMU 102, and a domain controller 103. The onboard camera 101 is equipped with an image sensor 1011. The IMU is an Inertial Measurement Unit.

[0038] The image sensor 1011 incorporates two types of pixel units with different light-sensing capabilities, enabling the vehicle-mounted camera 101 to acquire light signals from highlight areas and shadow areas in the environment surrounding the vehicle 100, respectively. These light signals are then converted into digitized electrical signals to obtain highlight pixel data and shadow pixel data. The IMU 102 is a set of electronic sensor components used to acquire the current motion parameters of the vehicle 100. The domain controller 103 receives pixel data from the vehicle-mounted camera 101 and motion parameters from the IMU 102, determines a fused image based on the motion parameters and pixel data, and provides the fused image to the autonomous driving system.

[0039] Figure 2 This is a flowchart of a fuzzy compensation method provided according to an embodiment of this application. The method is executed by the vehicle's domain controller. Figure 2 As shown, the fuzzy compensation method includes the following steps: S201, linearly fuses the highlight pixel data and dark pixel data collected by the image sensor of the vehicle-mounted camera to obtain the basic image.

[0040] In this embodiment of the application, a vehicle-mounted camera is installed in the vehicle. The vehicle-mounted camera collects highlight pixel data and dark pixel data of the surrounding scene through an image sensor. The domain controller obtains the highlight pixel data and dark pixel data sent by the vehicle-mounted camera and performs linear fusion on the highlight pixel data and dark pixel data to obtain a basic image.

[0041] The image sensor, a photosensitive chip inside the vehicle-mounted camera, converts light signals into digitized electrical signals to obtain pixel data. Highlight pixel data records the brightness levels and details of highlight areas; shadow pixel data records the brightness information of shadow areas. Linear fusion refers to a pixel-by-pixel weighted sum of highlight and shadow pixel data with fixed weights (e.g., 0.5 each). The base image contains structural information about the environment corresponding to the highlight and shadow pixel data, as well as motion blur information caused by vehicle movement, used for subsequent blur compensation of the highlight and shadow pixel data. Structural information reflects the spatial arrangement characteristics of object shape, edge sharpness, texture distribution, and texture clarity.

[0042] To acquire both highlight and shadow pixel data in a single exposure, the image sensor in this embodiment of the application is equipped with two types of pixel units with different light-sensing capabilities. Highlight pixel data is acquired using pixel units with weaker light-sensing capabilities, while shadow pixel data is acquired using pixel units with stronger light-sensing capabilities. A pixel unit is the smallest light-sensitive unit (i.e., a single pixel) on an image sensor, used to receive light at its corresponding location and convert its intensity into an electrical signal. Each pixel unit corresponds to one sampling point. Pixel units with different light-sensing capabilities have varying degrees of sensitivity to light signals. Within the same exposure time, pixel units with stronger light-sensing capabilities can generate stronger electrical signals in low light, while pixel units with weaker light-sensing capabilities are less prone to saturation in strong light, thus preserving highlight details.

[0043] S202, extracts sharp components from the base image based on the vehicle's motion parameters.

[0044] In this embodiment, while the image sensor acquires pixel data, the IMU measures the vehicle's motion parameters and sends these parameters to the domain controller. This allows the domain controller, after acquiring the base image, to extract the sharp component from the base image based on the vehicle's motion parameters during pixel data acquisition. The motion parameters describe the vehicle's motion state, including its linear velocity, angular velocity, and direction of motion, and are measured by the IMU within the vehicle and sent to the domain controller. Linear velocity is the vehicle's speed along its direction of travel; angular velocity is the angular velocity of the vehicle's rotation along its vertical axis. The IMU includes an accelerometer and a gyroscope. The accelerometer measures the linear velocity, and the gyroscope measures the angular velocity. The sharp component refers to the structural information in the base image that is not obscured by motion blur, reflecting the sharp edges and textures that should exist in the base image.

[0045] During vehicle operation, there is relative motion between the vehicle camera and the surrounding scene. This relative motion is determined by the vehicle's motion parameters. In other words, motion blur caused by relative motion is related to the motion parameters. Therefore, the blurred areas caused by vehicle motion in the base image can be determined by the motion parameters. Furthermore, the blurred components and sharp components can be extracted from the blurred areas.

[0046] S203, based on the sharpness component, performs blur compensation on the highlight pixel data and shadow pixel data, and determines the fused image based on the blurred highlight pixel data and shadow pixel data.

[0047] In this embodiment, after acquiring the sharp component, the domain controller uses it as a compensation basis to compensate for the blurred areas in the highlight and shadow pixel data, thereby determining the fused image based on the compensated highlight and shadow pixel data. Blur compensation refers to the process of repairing and replacing structural information in the highlight and shadow pixel data that has been contaminated by motion blur.

[0048] It should be noted that the fused image is obtained by fusing highlight pixel data and dark pixel data. In this embodiment, blur compensation can be achieved by adding a sharp component during the fusion process of highlight pixel data and dark pixel data.

[0049] This application provides a blur compensation method that extracts a sharp component, i.e., the original structural information that is not contaminated by motion blur, from a base image using motion parameters. The sharp component is then used to compensate for motion blur in both highlight and shadow pixel data. This results in a fused image that not only preserves details in both highlight and shadow areas but also restores edge structures and texture information contaminated by motion blur. This provides a reliable visual image for the perception and decision-making capabilities of autonomous driving systems, ensuring the effectiveness of autonomous driving systems and improving vehicle safety.

[0050] The above embodiments describe extracting the sharp component from a base image based on motion parameters, and then using the sharp component to perform blur compensation on highlight and shadow pixel data. Based on this, Figure 3 This is a flowchart of another blur compensation method provided according to an embodiment of this application. The method describes how to determine the motion blur intensity based on motion parameters, how to extract the sharp component based on the motion blur intensity, and how to perform blur compensation on highlight and shadow pixel data using the sharp component. This method is executed by the vehicle's domain controller. Figure 3 As shown, the fuzzy compensation method includes the following steps: S301 performs linear fusion of the highlight pixel data and dark pixel data collected by the image sensor of the vehicle-mounted camera to obtain the basic image.

[0051] Please refer to step S201 for the specific process, which will not be repeated here.

[0052] S302, determine the motion blur intensity of the base image based on the vehicle's motion parameters.

[0053] In this embodiment, motion blur is caused by vehicle motion. Therefore, the motion blur intensity is related to the vehicle's motion parameters. The motion blur intensity of the base image is determined based on the motion parameters. The motion blur intensity is used to quantify the degree of blurring of each pixel in the base image under vehicle motion.

[0054] In some embodiments, the blur length and blur direction of the base image are determined based on the vehicle's motion parameters; the base blur intensity of each pixel is determined based on the gradient magnitude of each pixel in the base image; and the base blur intensity is corrected based on the gradient direction, blur length, and blur direction of each pixel to obtain the motion blur intensity of each pixel.

[0055] In this context, blur length refers to the pixel displacement distance at the same spatial location in the surrounding scene captured by the image sensor within a single exposure due to vehicle movement; it is the length of the motion blur. Blur direction refers to the direction of pixel displacement at that spatial location, i.e., the direction of the motion blur, which is opposite to the direction of vehicle movement. The gradient magnitude of a pixel refers to the rate of change of its grayscale value in a spatial direction, reflecting the sharpness of the pixel's edge. The basic blur intensity is determined by the gradient magnitude; a larger gradient magnitude indicates more drastic local changes at the pixel and a smaller basic blur intensity, while a smaller gradient magnitude indicates smoother local changes at the pixel and a larger basic blur intensity, making the pixel more likely to belong to a smooth or blurred region. The gradient direction of a pixel is the direction in which its grayscale value changes the fastest.

[0056] After obtaining the motion parameters, the domain controller can substitute the motion parameters into Formula 1 to calculate the fuzzy length and fuzzy direction.

[0057] Formula 1: ; in, For fuzzy length; Let be the linear velocity of the vehicle. The angular velocity of the vehicle; This is the weighting coefficient for linear velocity. This is the weighting coefficient for angular velocity. and Used to balance the influence of linear velocity and angular velocity on fuzzy length; This refers to the single exposure duration of the vehicle-mounted camera, which is the time it takes for the image sensor to acquire pixel data. The larger the value, the greater the blur length caused by vehicle movement; The smaller the value, the smaller the fuzzy length caused by vehicle movement.

[0058] The gradient magnitude and direction of each pixel are related to its grayscale value. For each pixel, the Sobel operator (Sobel operator) is used to calculate the gradient components in the horizontal and vertical directions. The Sobel operator is a discrete difference operator, and its core principle is to use two sets of 3×3 convolution kernels to perform convolution operations on the pixel in the horizontal and vertical directions. The horizontal convolution operation is implemented using Equation 2 to obtain the horizontal gradient component; the vertical convolution operation is implemented using Equation 3 to obtain the vertical gradient component. Substituting the horizontal and vertical gradient components into Equation 4 yields the gradient magnitude of the pixel, and substituting them into Equation 5 yields the gradient direction of the pixel.

[0059] Formula 2: ; Formula 3: ; in, Indicates gradient components, subscript Indicates horizontal direction, subscript Indicates the vertical direction. Represents pixels The original grayscale value.

[0060] Formula 4: ; in, Represents pixels The gradient magnitude.

[0061] Formula 5: ; in, Represents pixels The gradient direction.

[0062] Substituting the gradient magnitude into Formula 6 yields the basic blur intensity of the pixel.

[0063] Formula Six: ; in, Represents pixels The basic fuzziness intensity, It is a stability constant to prevent the denominator from being zero.

[0064] Substituting the basic blur intensity, gradient direction, blur length, and blur direction into Formula 7 allows the gradient direction, blur length, and blur direction to correct the basic blur intensity, thus obtaining the motion blur intensity of the pixel.

[0065] Formula 7: ; in, Represents pixels The intensity of motion blur, The blur direction is defined. The first term in Formula 7 is the basic blur intensity; the second term is used to verify whether the low gradient of the pixel is caused by vehicle motion. Calculate the angle between the gradient direction and the blur direction. When the gradient direction and the blur direction are perpendicular, the low gradient of the pixel is caused by motion blur from vehicle movement; when the gradient direction and the blur direction are not perpendicular, the low gradient of the pixel originates from its own low texture. Used for adjustment The weights are assigned to ensure that areas with more severe motion blur are properly balanced. The higher the credibility, the better. and These are the fusion weights for the first and second terms, used to balance them. They can be equal or unequal.

[0066] It should be noted that due to the varying illumination intensity at different spatial locations in the surrounding scene, the gradient magnitudes of each pixel differ significantly, leading to substantial differences in motion blur intensity. Therefore, this embodiment of the application, after determining the motion blur intensity of each pixel, normalizes the motion blur intensity, mapping it uniformly to the interval [0,1], so that a normalized motion blur intensity with a unified dimension can be used in subsequent calculations. The normalization formula is shown in Formula 8: Formula 8: ; in, For pixels The normalized motion fuzz intensity, The maximum motion blur intensity of each pixel in the base image. This is the minimum motion blur intensity for each pixel in the base image. It should be noted that, after calculating the normalized motion blur intensity, all subsequent motion blur intensities used in this embodiment are normalized motion blur intensities.

[0067] S303 extracts the sharp component from the base image based on the motion blur intensity.

[0068] In this embodiment, based on the motion blur intensity of each pixel, the sharp component corresponding to each pixel is extracted from the base image. The sharp component refers to the structural information that is not contaminated by motion blur.

[0069] In some embodiments, the base image is frequency-domain transformed based on the motion blur intensity to obtain the image spectrum of the base image; the image spectrum is filtered to obtain the sharp component spectrum of the base image; and the sharp component spectrum is inversely transformed in the frequency domain to obtain the sharp component of each pixel in the base image.

[0070] Frequency domain transformation refers to converting the base image from the spatial domain (pixel coordinates) to the frequency domain (frequency coordinates) using Fourier transform, so that the structural information in the image is presented in the form of frequency components. The image spectrum is the representation of the base image after frequency domain transformation. The sharp component spectrum consists of the frequency components in the image spectrum that are not affected by motion blur. The sharp component is the structural information in the base image that is not affected by motion blur and can be used as a reference to compensate for blurred areas.

[0071] In some embodiments, the process of obtaining the image spectrum is as follows: based on the motion blur intensity of each pixel in the base image, the original gray values ​​of each pixel are weighted to obtain the blurred modulated gray values ​​of each pixel; Fourier transform is performed on the blurred modulated gray values ​​of all pixels in the base image to obtain the image spectrum of the base image. The blurred modulated gray values ​​represent the gray-level response after blur modulation of each pixel.

[0072] Substituting the normalized motion blur intensity and original grayscale value of each pixel into Formula 9 yields the blurred modulation grayscale value. Since the normalized motion blur intensity ranges from [0,1], pixels with higher motion blur intensity retain their original grayscale value, while pixels with lower motion blur intensity have their original grayscale value suppressed. A Fourier transform is then performed on the blurred modulation grayscale values ​​of all pixels using Formula 10. Because the sharp areas in the base image are suppressed, the energy of the image spectrum obtained after the Fourier transform primarily originates from the blurred areas caused by motion blur.

[0073] Formula Nine: ; Among them, subscript Indicates fuzzy modulation. Represents pixels The blurred modulated grayscale value.

[0074] Formula 10: ; in, Represents frequency coordinates. This represents the image spectrum obtained after performing a Fourier transform on the blurred modulated gray values, including frequency coordinates. Amplitude and phase information, The number of pixels in the horizontal direction of the base image. The number of pixels in the vertical direction of the base image. It is the imaginary unit. The horizontal coordinate in the frequency domain Horizontal coordinates in the spatial domain Normalized correlation terms, The vertical coordinate in the frequency domain perpendicular coordinates in the spatial domain The normalized correlation terms.

[0075] By transforming the frequency domain, motion blur is represented in the frequency domain by dark fringes along a specific direction, so that the sharp component spectrum and the blurred spectrum can be separated by filtering later.

[0076] In some embodiments, after determining the blur length and blur direction of the base image, a directional filtering function is constructed based on the blur length and blur direction. When filtering the image spectrum, the image spectrum is directionally filtered based on the directional filtering function to obtain the sharp component spectrum of the base image.

[0077] The directional filtering function is constructed based on Formula 11, and the clear component spectrum is extracted based on Formula 12.

[0078] Formula 11: ; in, This indicates the directional filter function at a frequency point. The value of is in the range [0,1]. Within this range, The larger the value, the higher the degree of preservation of that frequency point. The smaller the value, the stronger the suppression at that frequency point. Representing frequency point The larger the value of the projection component in the blur direction, the closer the frequency point is to the blur direction, the higher the degree of influence of blur, and the more it should be suppressed. The normalization factor represents the fuzziness length and is used to normalize the projected components. The larger the fuzziness length, the wider the suppressed frequency band. Parameters for controlling the filtering strength, The larger the value, the stronger the inhibitory effect. This is used to map the values ​​within parentheses to the interval [0,1] to achieve smooth filtering.

[0079] Formula 12: ; in, Representing frequency point The clear component spectrum. Used to represent the frequency components in the ambiguity direction, which are related to The product of represents the spectrum of the fuzzy components, therefore Used to represent frequency components that deviate from the ambiguity direction, which are related to The product of these components represents the clear component spectrum.

[0080] After obtaining the sharp component spectrum, the sharp component spectrum is inversely transformed in the frequency domain using Formula Thirteen to obtain the sharp component of each pixel.

[0081] Formula Thirteen: ; in, Represents pixels The clear proportions.

[0082] It should be noted that, in this embodiment of the application, a frequency domain separation model can also be integrated into the domain controller. The sharp component is extracted through the frequency domain separation model. After obtaining the base image and the normalized motion blur intensity of each pixel, the base image and the normalized motion blur intensity obtained in step S302 are input into the frequency domain separation model, causing the frequency domain separation model to output the sharp component. When training the frequency domain separation model, sharp images without motion blur and blurred images containing motion blur, acquired at the same time in the same scene, are used as training datasets. The blurred image is input into the frequency domain separation model to obtain the sharp component output by the model. Then, the sharp component and the sharp image are input into the loss function to obtain the difference value between them. Finally, the parameters in the directional filtering function are adjusted based on the difference value. Iterative updates are performed to minimize this difference, ensuring that the sharp component spectrum obtained through the directional filtering function, after inverse frequency domain transformation, yields a sharp component as close as possible to the sharp image. The loss function is constructed as shown in Formula Fourteen: Formula Fourteen: ; in, Represents the pixel points output by the frequency domain separation model. The difference between the sharp component and the original grayscale value. Represents pixels in a clear image The original grayscale value.

[0083] S304 fuses the sharp component, highlight pixel data, and shadow pixel data to obtain a fused image.

[0084] In this embodiment, a sharpness component is introduced during the fusion of highlight pixel data and dark pixel data. The three are fused pixel by pixel to compensate for the blur of the highlight pixel data and dark pixel data through the sharpness component, so as to obtain a deblurred fused image.

[0085] In some embodiments, after obtaining the sharp component of each pixel, the motion blur intensity of each pixel is determined as the first fusion weight of the sharp component of each pixel. When fusing the sharp component, highlight pixel data, and shadow pixel data, a second fusion weight of the first gray value of each pixel in the highlight pixel data is obtained, and a third fusion weight of the second gray value of each pixel in the shadow pixel data is obtained. Based on the first fusion weight, the second fusion weight, and the third fusion weight, the sharp component, the first gray value, and the second gray value of each pixel are weighted and fused to obtain each fused pixel. A fused image is generated based on each fused pixel.

[0086] During the fusion process, for each pixel, the normalized motion blur intensity of that pixel is determined as the first fusion weight of the sharp component. Pixels with higher normalized motion blur intensity also have a higher first fusion weight of the sharp component to repair the structural information destroyed by motion blur; pixels with lower normalized motion blur intensity have a lower first fusion weight of the sharp component to ensure that highlight and shadow pixel data dominate the fusion process, thereby preserving the original brightness and color of the pixel. The pixels in the highlight and shadow pixel data are in one-to-one correspondence. Before fusion, the second fusion weight of the first grayscale value and the third fusion weight of the second grayscale value of each pixel are first determined. During the fusion process, the grayscale value of each pixel is weighted according to the fusion weight.

[0087] After determining the first fusion weight of the sharp component, the second fusion weight of the first gray value, and the third fusion weight of the second gray value, the sharp component, the first gray value, and the second gray value are fused based on Formula 15 to obtain each fused pixel.

[0088] Formula 15: ; in, For pixels The corresponding fused pixels, As the first fusion weight, , As the second fusion weight, As the third fusion weight, For pixels The first grayscale value in the highlight pixel data For pixels The second grayscale value in the dark pixel data.

[0089] In some embodiments, the process of obtaining the second fusion weight and the third fusion weight is as follows: obtaining the basic fusion weight of the first gray value of each pixel in the highlight pixel data, obtaining the basic fusion weight of the second gray value of each pixel in the dark pixel data; and correcting the basic fusion weight of the first gray value and the basic fusion weight of the second gray value based on the motion blur intensity and sharpness component of each pixel, respectively, to obtain the second fusion weight of the first gray value and the third fusion weight of the second gray value.

[0090] The basic fusion weight is determined based on the first or second grayscale value of each pixel. However, this basic fusion weight does not consider whether the first or second grayscale value is affected by motion blur. Therefore, it needs to be adjusted based on the motion blur intensity and the sharpness component. Both motion blur intensity and the sharpness component are used to determine the reliability of the first or second grayscale value. The higher the motion blur intensity, the lower the basic fusion weight of the first and second grayscale values ​​should be. The local mean of the sharpness component of each pixel is calculated using the sharpness components of each pixel and the sharpness components of a preset number of surrounding pixels. This local mean reflects the grayscale value that the pixel should present after deblurring. Therefore, the local mean of the sharpness component is compared with the first and second grayscale values ​​to further determine their reliability. The larger the difference between the local mean of the sharpness component and the first or second grayscale value, the lower the basic fusion weight of the first or second grayscale value should be.

[0091] In some embodiments, the process of obtaining the basic fusion weights is as follows: based on the saturation response value of the image sensor, the first gray value of each pixel in the highlight pixel data and the second gray value in the dark pixel data are normalized to obtain the normalized gray value of each pixel; based on the normalized gray value of each pixel, the basic fusion weight of the first gray value and the basic fusion weight of the second gray value are determined.

[0092] The saturation response value is the maximum grayscale value that the image sensor can output, determined by the bit depth of the image sensor. After determining the first and second grayscale values ​​of each pixel, the first grayscale value, the second grayscale value, and the saturation response value are substituted into Formula Sixteen to obtain the normalized grayscale value of the pixel. Through the normalized grayscale value, the grayscale values ​​of all pixels are quantized to a uniform scale of 0-1. At the same time, the normalized grayscale value is used as the basis for assigning basic fusion weights to each pixel. When the normalized grayscale value approaches 0, it indicates that the pixel is in a dark area, and the basic fusion weight of the second grayscale value is greater than that of the first grayscale value. When the normalized grayscale value approaches 1, it indicates that the pixel is in a highlight area, and the basic fusion weight of the first grayscale value is greater than that of the second grayscale value. When the normalized grayscale value approaches an intermediate value between 0 and 1, the basic fusion weights of the first and second grayscale values ​​tend to be balanced.

[0093] Formula Sixteen: ; in, For pixels The normalized gray value, This represents the saturation response value of the image sensor.

[0094] It should be noted that, in this embodiment, a weight prediction model can also be integrated into the domain controller, outputting a second fusion weight and a third fusion weight. When training the weight prediction model, a training dataset is first constructed based on highlight pixel data, shadow pixel data, and HDR images synchronously acquired by professional equipment in the same scene at the same time. For each set of sample data in the training dataset, the sharp component, normalized grayscale value, and normalized motion blur intensity of each pixel are determined based on the highlight pixel data and shadow pixel data, thereby determining the local mean of the sharp component of each pixel. For each pixel, a three-dimensional feature vector is determined based on the normalized grayscale value, the local mean of the sharp component, and the normalized motion blur intensity. The three-dimensional feature vector of each pixel is used as the input to the weight prediction model, causing the weight prediction model to output a second fusion weight based on the first grayscale value and a third fusion weight based on the second grayscale value of each pixel. The normalized motion blur intensity of each pixel is used as the first fusion weight of the sharp component. Based on the first fusion weight, the second fusion weight, and the third fusion weight, the sharp component, the first gray value, and the second gray value are fused. The loss value between the gray value of each pixel in the HDR image and the corresponding fusion result is calculated. With the goal of minimizing the loss value, the model parameters of the weight prediction model are iteratively updated through the backpropagation algorithm until the loss value converges or reaches the preset training rounds, and the trained weight prediction model is obtained.

[0095] This application provides a blur compensation method that determines the motion blur intensity of each pixel through motion parameters, preventing misjudgments of low-texture areas based solely on image features and improving the detection accuracy of blurred areas caused by motion blur. By performing frequency domain transformation on the base image, the frequency components contaminated by motion blur and those not contaminated can be accurately separated in the frequency domain. This makes the extraction of the sharp component directionally selective, ensuring the accuracy of sharp component extraction and avoiding ringing artifacts caused by deblurring in the spatial domain. By fusing the sharp component, highlight pixel data, and dark pixel data, the sharp component compensates for the motion blur in the highlight and dark pixel data. The resulting fused image not only preserves the details of both highlight and dark areas but also restores the edge structure and texture information contaminated by motion blur. This provides a reliable visual image for the perception and decision-making capabilities of autonomous driving systems, ensuring the effectiveness of autonomous driving systems and improving vehicle safety. The motion blur intensity is determined as the fusion weight of the sharp component, so that the compensation intensity of the sharp component matches the degree to which the pixel is affected by motion blur; the more severe the blur, the more compensation is given, and the less severe the blur, the less compensation is given. When determining the fusion weights of the first and second grayscale values, a basic fusion weight is first determined based on the normalized grayscale value. For pixels in dark areas, the basic fusion weight of the second grayscale value is set greater than that of the first grayscale value. For pixels in highlight areas, the basic fusion weight of the first grayscale value is set greater than that of the second grayscale value. This avoids underexposure or overexposure caused by fixing weights in either dark or highlight areas. The basic fusion weights are then adjusted based on the motion blur intensity and sharpness component. In sharp areas, the original grayscale value is preserved to avoid loss of detail. In blurred areas, the basic fusion weights of the first and second grayscale values ​​are reduced to suppress the contribution of blurred areas.

[0096] Figure 4 This is a block diagram of a fuzzy compensation device according to an embodiment of this application. The device is used to perform the steps of the above-described fuzzy compensation method, see below. Figure 4 The device includes: The linear fusion unit 401 is configured to linearly fuse the highlight pixel data and dark pixel data collected by the image sensor of the vehicle-mounted camera to obtain a base image. The component extraction unit 402 is configured to extract sharp components from the base image based on the vehicle's motion parameters; The blur compensation unit 403 is configured to perform blur compensation on the highlight pixel data and the shadow pixel data based on the sharp component, so as to determine the fused image based on the blur-compensated highlight pixel data and shadow pixel data.

[0097] In some embodiments, the component extraction unit 402 includes: a blur intensity determination subunit, configured to determine the motion blur intensity of the base image based on the motion parameters of the vehicle; The sharp component extraction subunit is configured to extract sharp components from the base image based on motion blur intensity.

[0098] In some embodiments, the blur intensity determination subunit is configured to determine the blur length and blur direction of the base image based on the vehicle's motion parameters; The basic blur intensity of each pixel is determined based on the gradient magnitude of each pixel in the base image. Based on the gradient direction, blur length, and blur direction of each pixel, the basic blur intensity is corrected to obtain the motion blur intensity of each pixel.

[0099] In some embodiments, the sharp component extraction subunit is configured to perform frequency domain transformation on the base image based on the motion blur intensity to obtain the image spectrum of the base image; Filter the image spectrum to obtain the sharp component spectrum of the base image; The sharp component spectrum is inversely transformed in the frequency domain to obtain the sharp component of each pixel in the base image.

[0100] In some embodiments, the sharp component extraction subunit is configured to weight the original gray value of each pixel based on the motion blur intensity of each pixel in the base image to obtain the blurred modulated gray value of each pixel. The image spectrum of the base image is obtained by performing a Fourier transform on the blurred modulated gray values ​​of all pixels in the base image.

[0101] In some embodiments, the fuzziness intensity determination subunit is configured to construct a directional filtering function based on the fuzziness length and the fuzziness direction; The sharp component extraction subunit is configured to perform directional filtering on the image spectrum based on a directional filtering function to obtain the sharp component spectrum of the base image.

[0102] In some embodiments, the blur compensation unit 403 includes a fusion subunit configured to fuse the sharp component, highlight pixel data and dark pixel data to obtain a fused image.

[0103] In some embodiments, the sharp component extraction subunit is configured to determine the motion blur intensity of each pixel as the first fusion weight of the sharp component of each pixel. The fusion subunit is configured to obtain a second fusion weight for each pixel in the highlight pixel data and a third fusion weight for each pixel in the dark pixel data. Based on the first fusion weight, the second fusion weight, and the third fusion weight, the sharp component, the first gray value, and the second gray value of each pixel are weighted and fused to obtain each fused pixel. A fused image is generated based on each fused pixel.

[0104] In some embodiments, the fusion subunit is configured to obtain the basic fusion weight of the first gray value of each pixel in the highlight pixel data and the basic fusion weight of the second gray value of each pixel in the dark pixel data. Based on the motion blur intensity and sharpness component of each pixel, the basic fusion weights of the first grayscale value and the second grayscale value are modified respectively to obtain the second fusion weight of the first grayscale value and the third fusion weight of the second grayscale value.

[0105] In some embodiments, the fusion subunit is configured to normalize the first gray value of each pixel in the highlight pixel data and the second gray value in the dark pixel data according to the saturation response value of the image sensor, so as to obtain the normalized gray value of each pixel. Based on the normalized gray values ​​of each pixel, the basic fusion weights of the first gray value and the second gray value are determined.

[0106] This application provides a blur compensation device that extracts a sharp component, i.e., the original structural information that is not contaminated by motion blur, from a base image using motion parameters. The sharp component is used to compensate for motion blur in highlight and shadow pixel data, so that the final fused image can not only take into account the details of highlight and shadow areas, but also restore the edge structure and texture information contaminated by motion blur. This provides a reliable visual image for the perception and decision-making capabilities of autonomous driving systems, ensuring the effectiveness of autonomous driving systems and improving vehicle safety.

[0107] It should be noted that the fuzzy compensation device provided in the above embodiments is only illustrated by the division of the above functional modules when running the application. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the fuzzy compensation device and the fuzzy compensation method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0108] Figure 5 This is a schematic diagram of the structure of a domain controller according to an embodiment of this application.

[0109] Typically, the domain controller 103 includes: a main control module 1031, a CAN (Controller Area Network) interface 1032, a hardwired input interface 1033, and a hardwired output interface 1034. The main control module 1031 is connected to the CAN interface 1032, the hardwired input interface 1033, and the hardwired output interface 1034, respectively.

[0110] The main control module 1031 typically includes a processor and memory. The processor may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor can be implemented using at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), or PLA (Programmable Logic Array). The processor may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content required to be displayed on the vehicle's screen. In some embodiments, the processor may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning. The memory may include one or more computer-readable storage media, which may be non-transitory. The memory may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, a non-transitory computer-readable storage medium in the memory is used to store at least one computer program, which is executed by a processor to implement the fuzzy compensation method provided in the method embodiments of this application.

[0111] The CAN interface 1032 may include a powertrain CAN interface, a motor CAN interface, and a diagnostic CAN interface. The powertrain CAN interface is used to communicate with the vehicle's powertrain module, the motor CAN interface is used to communicate with the vehicle's motor controller, and the diagnostic CAN interface is used to communicate with diagnostic equipment.

[0112] The hard-wired input interface 1033 is used to receive hard-wired control signals. The hard-wired output interface 1034 is used to send control commands to the vehicle's electronic control components, causing the vehicle's electronic control components to perform corresponding actions. The vehicle's electronic control components include a power management system, a motor controller, an on-board charger, and a body control system.

[0113] The main control module 1031 can communicate with the vehicle's powertrain module, motor controller, and diagnostic equipment via the CAN interface 1032, and generate control commands based on the hard-wired control signals received by the hard-wired input interface 1033, so as to send the control commands to the vehicle's electronic control components via the hard-wired output interface 1034.

[0114] Those skilled in the art will understand that Figure 5 The structure shown does not constitute a limitation on the domain controller 103 and may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0115] This application also provides a computer-readable storage medium storing at least one computer program, which is loaded and executed by a domain controller's processor to implement the operations performed by the domain controller in the fuzzy compensation method of the above embodiments. For example, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, or an optical data storage device, etc.

[0116] This application also provides a computer program product, including a computer program loaded and executed by a processor to implement the fuzzy compensation method as described in the above embodiments.

[0117] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0118] The above are merely optional embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A fuzzy compensation method, characterized in that, The method includes: The highlight pixel data and dark pixel data collected by the image sensor of the vehicle's on-board camera are linearly fused to obtain the base image; Based on the vehicle's motion parameters, a sharp component is extracted from the base image; Based on the sharpness component, blur compensation is performed on the highlight pixel data and the shadow pixel data to determine the fused image based on the blurred and compensated highlight pixel data and shadow pixel data.

2. The method according to claim 1, characterized in that, The extraction of sharp components from the base image based on the vehicle's motion parameters includes: The motion blur intensity of the base image is determined based on the vehicle's motion parameters; The sharp component is extracted from the base image based on the motion blur intensity.

3. The method according to claim 2, characterized in that, Determining the motion blur intensity of the base image based on the vehicle's motion parameters includes: Based on the vehicle's motion parameters, the blur length and blur direction of the base image are determined; Based on the gradient magnitude of each pixel in the base image, the basic blur intensity of each pixel is determined; Based on the gradient direction, blur length, and blur direction of each pixel, the basic blur intensity is corrected to obtain the motion blur intensity of each pixel.

4. The method according to claim 3, characterized in that, The extraction of the sharp component from the base image based on the motion blur intensity includes: Based on the motion blur intensity, the base image is transformed in the frequency domain to obtain the image spectrum of the base image; The image spectrum is filtered to obtain the sharp component spectrum of the base image; The sharp component spectrum is inversely transformed in the frequency domain to obtain the sharp component of each pixel in the base image.

5. The method according to claim 4, characterized in that, The step of performing frequency domain transformation on the base image based on the motion blur intensity to obtain the image spectrum of the base image includes: Based on the motion blur intensity of each pixel in the base image, the original grayscale value of each pixel is weighted to obtain the blurred modulated grayscale value of each pixel. The image spectrum of the base image is obtained by performing a Fourier transform on the blurred modulated gray values ​​of all pixels in the base image.

6. The method according to claim 4, characterized in that, After determining the blur length and blur direction of the base image based on the vehicle's motion parameters, the method further includes: A directional filtering function is constructed based on the fuzz length and the fuzz direction; The step of filtering the image spectrum to obtain the sharp component spectrum of the base image includes: The image spectrum is directionally filtered based on the directional filtering function to obtain the sharp component spectrum of the base image.

7. The method according to claim 4, characterized in that, The step of performing blur compensation on the highlight pixel data and the shadow pixel data based on the sharp component, and determining the fused image based on the blurred-compensated highlight pixel data and shadow pixel data, includes: The sharp component, the highlight pixel data, and the dark pixel data are fused to obtain a fused image.

8. The method according to claim 7, characterized in that, After performing inverse frequency domain transformation on the sharp component spectrum to obtain the sharp components of each pixel in the base image, the method further includes: The motion blur intensity of each pixel is determined as the first fusion weight of the sharp component of each pixel; The process of fusing the sharp component, the highlight pixel data, and the dark pixel data to obtain a fused image includes: A second fusion weight is obtained for the first gray value of each pixel in the highlight pixel data, and a third fusion weight is obtained for the second gray value of each pixel in the dark pixel data; Based on the first fusion weight, the second fusion weight, and the third fusion weight, the sharp component, the first gray value, and the second gray value of each pixel are weighted and fused to obtain each fused pixel. A fused image is generated based on each of the fused pixels.

9. The method according to claim 8, characterized in that, The step of obtaining a second fusion weight for the first grayscale value of each pixel in the highlight pixel data and a third fusion weight for the second grayscale value of each pixel in the dark pixel data includes: Obtain the basic fusion weight of the first gray value of each pixel in the highlight pixel data, and obtain the basic fusion weight of the second gray value of each pixel in the dark pixel data; Based on the motion blur intensity and sharpness component of each pixel, the basic fusion weights of the first grayscale value and the second grayscale value are modified respectively to obtain the second fusion weight of the first grayscale value and the third fusion weight of the second grayscale value.

10. The method according to claim 9, characterized in that, The step of obtaining the basic fusion weights for the first grayscale values ​​of each pixel in the highlight pixel data and obtaining the basic fusion weights for the second grayscale values ​​of each pixel in the dark pixel data includes: Based on the saturation response value of the image sensor, the first gray value of each pixel in the highlight pixel data and the second gray value in the dark pixel data are normalized to obtain the normalized gray value of each pixel. Based on the normalized grayscale value of each pixel, the basic fusion weight of the first grayscale value and the basic fusion weight of the second grayscale value are determined.

11. A fuzzy compensation device, characterized in that, The device includes: The linear fusion unit is configured to linearly fuse the highlight pixel data and dark pixel data collected by the image sensor of the vehicle-mounted camera to obtain a base image. A component extraction unit is configured to extract sharp components from the base image based on the vehicle's motion parameters; The blur compensation unit is configured to perform blur compensation on the highlight pixel data and the shadow pixel data based on the sharp component, so as to determine the fused image based on the blur-compensated highlight pixel data and shadow pixel data.

12. The apparatus according to claim 11, characterized in that, The component extraction unit includes: A blur intensity determination subunit is configured to determine the motion blur intensity of the base image based on the motion parameters of the vehicle; A sharp component extraction subunit is configured to extract sharp components from the base image based on the motion blur intensity.

13. The apparatus according to claim 12, characterized in that, The blur intensity determination subunit is configured to determine the blur length and blur direction of the base image based on the motion parameters of the vehicle. Based on the gradient magnitude of each pixel in the base image, the basic blur intensity of each pixel is determined; Based on the gradient direction, blur length, and blur direction of each pixel, the basic blur intensity is corrected to obtain the motion blur intensity of each pixel.

14. The apparatus according to claim 13, characterized in that, The sharp component extraction subunit is configured to perform frequency domain transformation on the base image based on the motion blur intensity to obtain the image spectrum of the base image; The image spectrum is filtered to obtain the sharp component spectrum of the base image; The sharp component spectrum is inversely transformed in the frequency domain to obtain the sharp component of each pixel in the base image.

15. The apparatus according to claim 14, characterized in that, The sharp component extraction subunit is configured to weight the original grayscale value of each pixel based on the motion blur intensity of each pixel in the base image to obtain the blurred modulated grayscale value of each pixel. The image spectrum of the base image is obtained by performing a Fourier transform on the blurred modulated gray values ​​of all pixels in the base image.

16. The apparatus according to claim 14, characterized in that, The fuzz intensity determination subunit is configured to construct a directional filtering function based on the fuzz length and the fuzz direction; The sharp component extraction subunit is configured to perform directional filtering on the image spectrum based on the directional filtering function to obtain the sharp component spectrum of the base image.

17. The apparatus according to claim 14, characterized in that, The fuzzy compensation unit includes: The fusion subunit is configured to fuse the sharp component, the highlight pixel data, and the dark pixel data to obtain a fused image.

18. The apparatus according to claim 17, characterized in that, The sharp component extraction subunit is configured to determine the motion blur intensity of each pixel as the first fusion weight of the sharp component of each pixel. The fusion subunit is configured to obtain a second fusion weight for the first gray value of each pixel in the highlight pixel data, and a third fusion weight for the second gray value of each pixel in the dark pixel data. Based on the first fusion weight, the second fusion weight, and the third fusion weight, the sharp component, the first gray value, and the second gray value of each pixel are weighted and fused to obtain each fused pixel. A fused image is generated based on each of the fused pixels.

19. The apparatus according to claim 18, characterized in that, The fusion subunit is configured to obtain a basic fusion weight for the first gray value of each pixel in the highlight pixel data, and to obtain a basic fusion weight for the second gray value of each pixel in the dark pixel data. Based on the motion blur intensity and sharpness component of each pixel, the basic fusion weights of the first grayscale value and the second grayscale value are modified respectively to obtain the second fusion weight of the first grayscale value and the third fusion weight of the second grayscale value.

20. The apparatus according to claim 19, characterized in that, The fusion subunit is configured to normalize the first gray value of each pixel in the highlight pixel data and the second gray value in the dark pixel data according to the saturation response value of the image sensor, so as to obtain the normalized gray value of each pixel. Based on the normalized grayscale value of each pixel, the basic fusion weight of the first grayscale value and the basic fusion weight of the second grayscale value are determined.

21. A domain controller, characterized in that, The domain controller includes a main control module, which includes a processor and a memory. The memory is used to store at least one computer program, which is loaded by the processor and executed according to any one of claims 1 to 10.

22. A vehicle, characterized in that, The vehicle includes a domain controller for loading and executing the fuzzy compensation method according to any one of claims 1 to 10.

23. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store at least one computer program, which is used to perform the fuzzy compensation method according to any one of claims 1 to 10.

Images